Implement mixed precision: f32 stencil, f64 pedestal

Browse Source

- Stencil arithmetic and shared memory use float (COMPUTE_TYPE alias).
- Pedestal accumulation stays double to preserve variance accuracy.

Notes:
- On RTX 4090, FP32 throughput is ~64× higher than FP64, so moving
  stencil math to float improves performance.
- Using float also avoids shared memory bank conflicts: stride-18 maps
  to distinct banks for 32-bit values, but caused conflicts with 64-bit.

...

This commit is contained in:

kferjaoui

2026-04-27 14:56:40 +02:00

parent 40f08fad92

commit ac96d1f688

3 changed files with 112 additions and 103 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

python/tests/ClusterFinderCUDA.ipynb

+52 -52

File diff suppressed because one or more lines are too long